finding
active
finding:trackeragent-trueskill-28-7-3-6-53-6-win-rateTrackerAgent TrueSkill μ=28.7 ± 3.6, 53.6% win rate
Best code agent outperforming six of seven LLMs.
Source paper
extracted_from(2026) · Robert Müller · Clemens Müller
Neighborhood — ranked by edge-count
Claims (2)
claim
- Calibration that conditional logic can beat cost-efficient LLMs in this setting.
- Ordering of code agents' strengths reflects key strategic dimensions.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- second-highest TrueSkill rating
- Second-best LLM, competitive with TrackerAgent.
- TrackerAgent won more than half its games
- Mid-field performance with larger uncertainty due to small sample.
- fourth-highest TrueSkill rating
- EconomyAgent TrueSkill rating
- TrackerAgent uses card counting to achieve high rating, a capability no LLM replicates
- TrackerAgent won over half of the combined-comp1 games.